Purpose: Exit dosimetry images are automatically acquired on the Halcyon™ linear accelerator for all treatment beams and could potentially be used to monitor daily treatment. We investigate the correlation between image analysis metrics calculated within the Portal Dosimetry Scripting API (PDSAPI) (version 15.1) and dosimetric changes throughout treatment.
Methods: Daily exit dosimetry images were compared to first-fraction exit images for 8 patients undergoing 16-fraction prone breast treatments. Dose-difference images, gamma analysis (2%/2mm and 3%/3mm, with 10% threshold) images, and gamma indices were calculated within the PDSAPI. For dose-difference images, a 10-70% pixel intensity region of interest (ROI) was selected and percent variations from the first-fraction pixel values were calculated; for daily 10-70% ROI images, an average dose-difference value was calculated. A 10-70% pixel threshold helped to focus on fluences traversed through breast tissue by eliminating out-of-field and out-of-breast dose. Dose-volume differences between the planned and daily delivered PTV parameters (dV90, dV95, dV100, dV103, dV105, and dV107) and Breast Evaluation PTV size were calculated on synthetic-CTs derived from daily-CBCT images. Pearson correlation coefficients with 0.05 significance levels were calculated between image analysis metrics and clinically-relevant DVH metrics.
Results: No gamma analysis metric produced significant correlations with any dose-volume parameters. Daily average dose-differences within ROIs and differences between -1-3% and +1-3% pixel deviations from dose-difference image metrics provided strong correlations (r < ±0.8) with deviation in DVH parameters (dV90, dV95, dV103, dV105, and dV107) normalized to planning values, as well as with breast volume changes.
Conclusion: Dose-difference tests between first and subsequent daily fraction exit portal dosimetry images could serve as a quick metric to identify overdose/underdose developments and changes in breast tissue volumes. These metrics can be easily implemented into clinical workflow using PDSAPI scripting capabilities for automated monitoring and alert for manual evaluation when large deviations are detected.